Claude lately: random downgrades, obvious nerfs, occasionally smart but consistently braindead, 5-hour limit burned by one small task, and rate-limits everywhere.
GPT 5.6 Sol just quietly gets the same work done. Slow, but relentless 🙂
K3 and GLM 5.2 are surging.
Cario Lee
@QCL15
Full-stack → AI | Building with LLMs | Vibe Coder → AI Builder
Making AI tools accessible